Gender Pay Equity: 15 Questions and Answers for You and Your Compensation Committee

Equity Methods Gender Pay Equity


At this year’s WorldatWork Total Rewards Conference in Dallas, I had the opportunity to participate in a panel discussion on gender pay equity. The session drew north of 400 people, showing just how important this topic is in our field.

Although we had a lot to talk about, we wanted to get the audience involved. So, we spent the first 20 minutes of the session polling the room for questions. Then we dedicated the rest of the time to answering them as we walked through a few prepared slides.

By popular demand—from attendees, that is—here’s an FAQ comprised of those questions. During the panel, there was not enough time to go into detail on each question, so this blog also gives me an opportunity to elaborate a bit more. (For a primer on gender pay equity, see this post.)

By the way, last year we published a similar—but much more detailed—FAQ on CEO pay ratio. If you think something like that would be helpful for gender pay equity, please let me know.

Gender Pay Equity at a High Level

1. How do you balance pay equity with performance?

At least in the United States, the idea of pay equity is integrally tied to ideas of performance and meritocracy. When we examine whether there’s a pay equity problem, what we’re really looking to do is validate that compensation is tied to an acceptable reason. Acceptable reasons include factors like tenure, performance, role, and location. Unacceptable reasons include someone’s gender, ethnicity, race, or some other attribute that doesn’t relate to their work and contribution.

Said differently, it’s okay if two people are paid differently because they do different work, or because one outperforms the other, or because they’re in different states and prevailing wages differ between those states. All those things have to do with the work the person is doing and not their identity.

When we study pay equity, we tend to look at performance, since different levels of performance should merit different levels of remuneration. The rub is that if a widespread gender bias exists, then this bias could show up even in performance ratings. Pay equity analyses are largely about disentangling messy cause-and-effect relationships of this nature: Is lower compensation due to genuinely weaker performance, or is a poor performance evaluation a cover, even subconsciously, for underlying biases? The good news is that with statistics and data modeling, you can attack the problem from multiple vantage points and form reasonable hypotheses as to what is actually taking place.

2. What does it mean to close the pay equity gap?

In its strictest sense, closing the pay equity gap means eliminating differences in pay that cannot be explained by appropriate reasons like role, location, performance, or tenure—providing equal pay for equal work. A pay equity gap exists when there are differences in pay not related to these factors, and further, one class of employees, commonly women, are disproportionately affected.

It’s important to understand that the issue centers on pay equity—today. But in the long run, it’s really about human capital. For example, pay levels might be fully explainable by appropriate factors and yet women or minorities are still underrepresented in leadership positions. This could be due to recruiting problems, promotion issues, skewed levels of attrition, or broader and more structural representation issues at the industry level. Either way, it’s worth understanding the gender, race, and ethnicity progression through the organizational hierarchy and where the unexplained failure points might be.

So today the focus is primarily pay equity. That’s a good place to begin, because it has more concrete data that we can study. But plan for the focus to broaden to overall human capital progression, in which pay equity plays a consistent part.

3. What about attributes besides gender, such as race and ethnicity?

Yes! This topic certainly goes beyond gender. Our presentation happened to focus on gender pay equity, but any potential sources of inequity should be studied. For instance, race is the second-most common factor to look at. What holds some companies back is that they don’t collect very much demographic information, so gender is all they’re able to look at.

Side note: There’s a growing school of thought around the idea of “intersectionality,” which looks at the unique challenges that (for example) black women face. The idea behind intersectionality is that there may be an even more nuanced layer of issues when you combine factors and look at them together instead of in isolation. Fortunately, statistical techniques exist to quite easily test whether there is a unique impact associated with intersectionality cases.

Defining the Importance of the Topic

4. Why is it important to correct pay equity gaps?

I really like devil’s-advocate questions. I’m sure the individual who asked this thinks pay equity is important, but wanted more specific reasons beyond simply, “It’s the right thing to do.”

In 1997, McKinsey coined the phrase “war for talent” to describe the emerging economy where companies’ abilities to acquire and retain top talent could be the defining factor in their success. Some 20 years later, these predictions are more acute than ever. Companies still compete by creating better semiconductor chips, better advertisements, better supply chains, or better Six Sigma processes. But it all seems secondary to getting their human capital right.

Getting human capital right helps companies improve on a number of dimensions, including not constricting the labor pool you draw from, ensuring growth and advancement for your top talent, and eliminating arbitrary causes of unnecessary turnover. Pay equity is an essential ingredient to keeping employees motivated. In a recent survey by Randstad US, 78% of employees said a workplace where all employees are treated equally is important to them. In short, we think pay equity is a linchpin to winning the war for talent, and as a result, a key to sustainability. And it’s the right thing to do.

5. How do we persuade top management that gender pay equity is important?

Among tech firms on the East or West Coast, pay equity is a hot-button issue to senior executives, investors, and boards. But that’s not universally true in other industries or geographies. Its importance might seem obvious, but I think pay equity needs thoughtful framing to convey the strategic relevance.

So how can you frame it? One way is as a tool in the war for talent. A pay equity study may reveal that your company:

  • Doesn’t have a pay equity problem (many companies don’t).
  • Doesn’t have a pay equity problem, but does have a related human capital problem (such as women dropping out of the workforce mid-career).
  • Does have a pay equity problem, but it’s not widespread or egregious (suggesting that it’s unintentional, solvable, and that overall compensation systems work well).
  • Does have a systemic pay equity problem.
  • Has poor representation of women or minorities at senior levels, or even in general.

The first finding is good news that you can share in your talent outreach. The second two findings provide an opportunity to address the issues so that they don’t undercut an otherwise highly effective talent strategy. Indeed, the data sets that reveal a problem can also hold clues about how you can address it. (More on that later.) The second to last finding is rare, but in the unlikely case it exists, identifying and managing the issues proactively will have a major talent benefit while reducing risk. The final problem is more common, and presents a distinct challenge for organizations. This is discussed later in this Q&A.

Another way to frame gender pay equity is through a risk management lens. Consider how organizations audit their information security and financial statements. Reasons include preemptively finding problems and being in a position to give positive assurances to external stakeholders. But a third reason is that should something bad happen, an audit can also show that the company made a good-faith effort to prevent it. The same can be true of pay equity. Even if a prior pay equity study had missed a situation in the data, the existence of the study is itself evidence that management took the matter seriously.

Finally, there’s the trend of compensation committees getting involved with human capital management. Companies are increasingly struggling with succession planning and personnel development at all levels of the organization. A robust pay equity analytics effort can equip senior management with answers when the compensation committee comes calling with questions.

To sum up, of course gender pay equity is about doing the right thing. At a very fundamental level, though, I think it’s about being proactive with the human capital assets of the organization and exercising good stewardship.

Measuring Pay Equity

6. How do you actually measure compensation differences to see if there is a bias?

Here’s how we approach the task at Equity Methods. We start with the hypothesis that compensation should be explainable based on factors like role, tenure, location, performance, education, and so on. We use a statistical technique called multiple regression that quantitatively explains the relationship between a dependent variable (pay) and a series of explanatory variables expected to predict compensation (e.g., role, performance, location, etc.). We also include what is called a “dummy variable” indicating gender, race, or any other area of interest. Dummy variable is a statistical term for the fact that it’s a binary (0/1) variable reflecting a certain trait so that we can test for the presence of systemic bias associated with that trait.

If pay equity exists, we will see no discernible impact of these dummy variables, and all of the dependent variability in compensation will be explained by the other variables. If, however, some of the variation is still predicted by our dummy, this indicates the possibility of systematic pay inequity. More often, however, we find there is not systemic bias but biases that are localized to smaller subgroups, such as individual business units or cost centers. The analysis then flags these groups to be looked at more closely.

By deploying this technique, the regression model can also be run to predict what compensation should be for each individual. For example, the model can be run to say that someone who is (for example) a band 9 vice president, working out of Atlanta, who has received above-average performance ratings and sits in the R&D group of the enterprise business unit, should be paid between $105,000 and $120,000. This model prediction is then used to identify whether any people fall outside the predicted band, and patterns in the outliers can be observed and analyzed.

If people or cohorts fall outside the model predictions, this doesn’t necessarily mean a bias exists. In fact, in an appropriate model, some percentage of employees will be paid outside the bands by construction. But it does mean that this particular model doesn’t explain why they’re paid what they are. These employees may be half men and half women, in which case you are probably fine; however, if women are five times as likely as men to be flagged as outliers, then a problem may exist. That’s why it’s necessary to use multiple models to get a more panoramic view of how compensation works. This is also why dialogue with business unit executives and HR generalists is important.

We like to think of models as ways to identify anomalies that need closer study, since not every dimension of pay strategy can be captured in the underlying data. For instance, variables like education, number of direct reports, and financial health of the cost center are not always readily available. But they could explain differences in pay.

7. How should jobs be aggregated for purposes of modeling? Should they be aggregated?

Before I answer this question, first let’s understand the context. Any serious pay equity analysis needs to look at the job or role a person performs. For example, software developers are usually paid more than business analysts. Job level notwithstanding, the underlying labor markets are different, which will result in different types of offers and pay mixes.

We can take that point to the extreme and say that in one sense, every single person has a distinct role. But that would be silly, since an analysis would fall apart without something to compare it to. So, we need a middle ground where we bundle together like employees while not taking it to the point that we’re analyzing fundamentally different roles together.

In general, statistical models work best when they can sift through large amounts of data in order to tease out nuanced relationships among variables. This is also why multivariate regression approaches work much better than calculating average pay for different groups. The regression model can include variables relating to role so that you gain the benefits of a large dataset without erasing key distinctions in the underlying data.

Also, the best pay equity processes are iterative. Modern computing power allows us to run advanced calculations on large datasets in next to no time at all. We develop multiple models, test them, and see how results converge or differ. As we do this, we assess the statistical efficacy of the different models. Where models show less statistical rigor than expected, we iterate to find an alternative specification that works better. Eventually, we have a suite of models that collectively yield the pay equity insight needed to begin forming conclusions.

In other words, there isn’t a hard-and-fast answer to how tightly jobs should be aggregated. Plan to try different levels of aggregation in order to find the right balance and what groupings make the most sense.

8. How do the approaches used in the US differ from pay equity reporting in the UK?

The Equality Act 2010 (Gender Pay Gap Information) Regulations 2017 in the United Kingdom require companies in Great Britain with over 250 employees to disclose certain gender pay gap information on their websites and a government website. The results are public and you can peruse them here.

The UK rules are incredibly prescriptive. The average and median male-to-female pay and bonus pay must be reported. The ratio of males to females who received a bonus must also be disclosed. Finally, companies are instructed to organize their workforce into four equal quartiles based on pay, and disclose the number of males and females in each quartile. Relatively specific definitions and protocols must be followed, and perhaps most importantly, this calculation is not at all robust to the fact that women and men perform different jobs within the organization. In fact, the results show that it is in many ways a better measure of the differences in roles than an indicator of equal pay for equal work.

As we’ve explained, pay equity processes in the US are much different because compensation committees and investors are the ones asking the questions. This leads companies to use state-of-the-art statistical techniques to holistically unpack the complexity of pay relationships. Disclosures, such as UK gender pay reporting, must be both formulaic and simplistic in order to apply across a wide range of companies.

Plan for questions as to why UK-reported results differ from those stemming from an analysis done in the home office. Since the home office project will generally be more rigorous and nuanced, part of its focus should be to explain the rationale behind any differences relative to UK-reported results.

Communication and Legal Privilege

9. How should pay equity processes be communicated internally?

So you’ve done a thoughtful pay equity analysis. What do you tell the organization? The right answer depends on your culture.

Some technology companies have such open cultures that the CEO responds to questions personally and is expected to be very open and transparent on even highly sensitive topics. However, in most cases, I’d say you should worry less about internal “marketing” and more on actual problem-solving. In our observations, public statements of the “We did X, which led to Y,” variety make the analysis more discoverable in any future litigation and start to feel like a PR campaign. But in some corporate contexts, this level of clarity is exactly what is needed.

This doesn’t suggest the right answer is pure silence, either, since shareholders and employees may be asking whether pay equity assessment processes are in place. But even then, basic messages like the following work well: “We absolutely look at pay equity and take the topic seriously. We have recurring processes to do that. Further, we also take preventative steps along the lines of anti-bias training for managers, workforce re-entry programs, and college recruiting initiatives to boost the diversity of entry-level hires.” Customize the specifics, but in our experience, phrasing like this seems to have more credibility while preserving the confidentiality of what is being done.

Regarding legal privilege, analyses like these generally should be commissioned by internal or external legal counsel as part of their effort to give legal advice to their client (the CEO, CHRO, or board). The reason maintaining privilege matters is because many cases won’t have black-and-white answers. As a result, organizations may require time to work through what the results mean and how to act on them. Contextually, a process like this is better kept under privilege than open to discovery should an exogenous lawsuit happen.

10. What is legal privilege and how does it play into things?

In our experience, different attorneys give slightly different viewpoints (again, we’re not attorneys). But the textbook explanation is as follows. In litigation, certain communications between a client and the client’s attorney are privileged (i.e., not discoverable by the opposing side) because they entail the client asking for legal advice. However, if the client then takes that privileged communication, forwards it to their colleague and initiates a separate discussion, then that separate discussion is almost certainly taking place outside the bounds of privilege.

In a pay equity study, generally the client asks their attorney to provide employment law support, of which pay equity is just a part. The attorney engages a quantitative specialist to develop robust statistical models and acts as a go-between for the results. The insights from those models help inform the attorney’s legal advice.

To be clear, many companies perform these studies outside the bounds of legal privilege. It’s a business decision to make based on your own organization’s circumstances and prior approaches to similar matters.

With or without legal privilege, some best practices apply. First, be careful about what you put in writing. Perhaps you see something in an analysis that frustrates you. In that moment, resist the temptation to send an email saying, “I can’t believe we did XYZ!” It’s never a bad time to pick up the phone and talk in person.

Second, tie up loose ends in your “work papers” (i.e., the files you keep on the study). A loose end would be an email or document that says something like the following without resolution: “We should probably correct the pay for these 10 people. What do you think?” Close out any hanging questions like that, or set a time to reassess it via an update to the files.

Finally, document the remediation steps you take. In the event of litigation, you need to show how you took the matter seriously by constantly initiating improvements to pay processes, training programs, and so on.

Remediating Pay Equity Problems

11. What happens if we detect a pay equity problem?

Remediation is an important topic. There are three broad approaches and, of course, many shades of gray in between.

The first approach is to communicate openly within the organization. This may come in a statement such as, “We performed a pay equity analysis and found no evidence of systematic bias. Further differences in pay were random between men and women, and most were easily explained by other factors. We made a total of $X in pay adjustments to 100 employees to remediate anomalous pay below expected levels.” This highly visible approach probably fits 15% to 20% of organizations.

Another way is to make pay adjustments so covertly that only a handful of people know the reason. Under this approach, a study yields suggested pay adjustments and those adjustments are woven into the next upcoming merit cycle, but without telling managers or HR leaders why. In most companies, the reasons behind pay adjustments aren’t fully transparent, which means it’s possible to boost pay adjustments without articulating why. This remediation strategy is typically seen in very large organizations.

The third approach, and usually our preferred one, is to take the results of an analysis to senior business line executives (or the HR generalists supporting them) and pull them into the dialogue. We call this the “Study, Consult, and Act” approach. It preserves discretion while yielding two useful benefits. For one thing, there may be factors that are relevant to the analysis but altogether missing from the data. They can help assess whether that’s the case. This outreach also sends a strong signal from the top that pay equity is a CEO-level priority.

We like the third approach because it’s sustainable. It gets people on board with the mission, marries the mathematical models with on-the-ground context, and opens an ongoing dialogue about pay equity. Making pay tweaks here and there is certainly important, but when done in isolation, it addresses symptoms and not causes.

12. How much should we budget for pay adjustments due to a pay equity problem?

This is an important question, since pay equity is important to every organization, but naturally many organizations have fixed budgets and might find it difficult to implement immediate corrective adjustments. Understandably so, there were skeptics in the room thinking: “It’s great that can shift budget money around. We probably can’t.”

The good news? Our expectation is that most cases won’t turn out to be budget-busters. That’s one reason why we believe more advanced statistical approaches are necessary to navigate the complexity of pay relationships and present a “measure twice, cut once” answer.

A side benefit of the Study, Consult, and Act approach is that senior management gains early indicators of potential pay biases. This way, if it looks like there will need to be pay adjustments, a dialogue can occur that allows affected parties to begin planning.

In terms of the chronology, a study usually takes six to eight weeks, at which point it’s possible to share how numbers are trending. The socialization process with business line executives usually takes another two or three months, since here the goal is showing them the results so that they can discreetly conduct internal research. After that, it’s time for business line executives to share their perspectives and senior management to make their decisions. All in all, the aim is to not let potential issues linger but to drive a methodical process that creates de facto training to business line executives. The byproduct is that the finance function can have time to digest the financial implications and adjust their budgets.

These processes work only when senior management and the board support them.

Nuances in a Pay Equity Study

13. When doing an analysis, how do you address roles that are predominantly occupied by men?

In our opinion, the starting point is a discussion around why these roles are predominantly occupied by men in the first place.

Take software engineering, a field in which studies suggest the percentage of females is 10% to 15%. The question to ask is whether there’s any valid reason for this. Most would say there isn’t.

Many leading companies have taken these statistics and used them to support overhauls to their recruiting procedures. For example, one high-tech company we work with appointed senior officers to forge relationships with local high schools and universities, creating awareness and excitement among women and minorities about careers in technology. In addition to doing good, they also positioned their organization to be at the forefront of future recruiting.

Of course, there is also the topic of self-selection, such as the assertion that women may simply prefer not to work on an oil rig. Be careful here, since one can easily counter-argue that perhaps the entire reason we don’t see many women working in oil rigs is the presence of structural biases that permeate society. Our suggestion is to devote time to internal dialogue on the topic of representation and your human capital strategy. Perhaps the answer is to show up at the local high schools and begin deconstructing stereotypes that lie behind current representation skews. At any rate, we at least want to raise the concept of self-selection as one that merits further discussion.

Smaller organizations may not have the resources to do what this particular organization did, but that doesn’t mean they’re without options. I’ll use Equity Methods as an example (we have just south of 100 professionals). By overhauling our approaches to campus and experienced-hire recruiting, we’ve significantly leapfrogged the male-female ratio seen in most consulting organizations while also achieving strong ethnic diversity.

What about data that shows women or minorities bailing out at higher rates once they hit a certain level in the organization? Such observations can inform improvements to internal mentoring programs, flex-time tracks, and workforce re-entry processes.

The point is, a multivariate regression model or any cohort analysis might perform worse when comparing the pay of hundreds of men to a handful of women. Nicely-sized datasets are the fuel these models run on. But in these cases where the gender imbalance is high and the model robustness is limited, this in and of itself lends insight and helps focus energy on strategies beyond just compensation.

14. When doing an analysis, how do you think about executives and are they treated differently?

Here I need to give the consultant’s notorious “it depends” answer. We like to include everyone in the analysis. Where we go from there depends on the dialogue and the data.

It’s not unheard of to have pay equity challenges even at executive levels, which we would generally define as the firm’s top 10% in terms of compensation. However, there are generally more unique considerations that need to be looked at and which are not in the data. For example, two business unit executives may have the same band level, live in the same state, and have equal performance ratings—but one earns much more because she manages a considerably larger P&L. If that particular fact isn’t in the HRIS data, a regression model won’t pick it up. Further, as there are fewer employees at each level, the models used lack the power to detect systematic bias.

Another factor with executives is that when problems exist, they’re more generally problems of representation. As a result, the study may trigger a more concentrated focus on helping women or minorities to progress through the career track (as I explain above).

Still, the power of modern computing allows analyses to be sliced multiple ways, so we would suggest including the full population and then being sure to cut the analytics by seniority level to see whether the story differs.

15. It’s not a secret that many women exit the workforce when they have children. How are these events handled in an analysis?

It’s important to start by defining the problem. One way of framing it is that talent leaves because they don’t think it’s possible to excel at work and at raising children at the same time. Another is that talent may wish to re-enter at some point, but it’s not clear how to make this easy and seamless.

However you define it, the first step in solving the problem is to study your data to understand what exactly is taking place. That way, conversations about strategy are grounded in facts. Suppose the data shows a clear trend of women exiting at a certain pay band and age level. What’s the right business response? We know some companies have created more part-time and flex-time roles so that they can help people keep one foot in the pond. This approach of course is easier said than done, since you can also end up with pay equity problems in more customized part-time roles.

Other companies have responded by changing their maternity leave policies so that women don’t feel like they’re forced into a choice at such a pivotal life event. Some companies are also extending paternity leave.

Your company may not in be a position to make wholesale changes to parental leave policies. Even so, you could examine the feasibility of part-time or flex-time opportunities. It’s also worth evaluating workforce re-entry programs, given how many women reach a stage where they do want to come back to work (full-time or part-time) and struggle to make that transition. From a human capital perspective, it makes all the sense in the world to understand how pockets of the labor market are being crowded out, making it harder to compete in the war for talent.


I hope you found this discussion helpful. If you’re among those who asked for this writeup, I’ll be sure to follow up personally.

I mentioned before that we published a more in-depth FAQ on CEO pay ratio. In that publication, we examined CEO pay ratio in a fair amount of detail. Do you think that gender pay equity merits a similar type of publication? If so, what would you like to cover? We think the broader topic of pay equity (extending even beyond gender) is considerably more complicated and meaningful than CEO pay ratio, and we’d like to help advance the dialogue in the industry. Please let me know what you think.